Training-Free Efficient Video Generation via Dynamic Token Carving

Supplementary Materials

Video are compressed with ffmpeg to reduce the size of the files.

Teaser Video

Hover on the video to see corresponding text prompts

A cat is wagging its tail.

HunyuanVideo-I2V / 338s / 4.43×

An Asian man with short hair in black tactical uniform and white clothes waves a firework stick.

HunyuanVideo-I2V / 338s / 4.43×

Several giant wooly mammoths approach treading through a snowy meadow, their long wooly fur lightly blows in the wind as they walk, snow covered trees and dramatic snow capped mountains in the distance, mid afternoon light with wispy clouds and a sun high in the distance creates a warm glow, the low camera view is stunning capturing the large furry mammal with beautiful photography, depth of field.

HunyuanVideo / 225s / 7.22×

The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it's tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.

HunyuanVideo-8GPU / 39s / 41.57×

A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about.

AccVideo / 76s / 2.12×

A cat is running.

Wan2.1-1.3B / 24s / 4.79×

More Showcases

Hover on the video to see corresponding text prompts

A woman with green hair smiling for the camera

Jenga+HunyuanI2V / 338s

A close up of leaves with water droplets on them

Jenga+HunyuanI2V / 338s

A woman smiles while holding a yellow flower

Jenga+HunyuanI2V / 338s

A woman in a wetsuit is swimming in the ocean

Jenga-3Stage / 157s

A young man with long, flowing hair sits on a rustic wooden stool in a cozy, dimly lit room, strumming an acoustic guitar. He wears a vintage denim jacket over a white t-shirt and faded jeans, his fingers skillfully moving across the strings. The warm glow of a nearby lamp casts soft shadows, highlighting his focused expression. As he plays, the camera captures close-ups of his hands, revealing intricate fingerpicking techniques. The room is adorned with musical memorabilia, including vinyl records and posters, creating an intimate, nostalgic atmosphere. His soulful performance resonates, filling the space with melodic harmony.

Jenga-3Stage / 157s

This close-up shot of a Victoria crowned pigeon showcases its striking blue plumage and red chest. Its crest is made of delicate, lacy feathers, while its eye is a striking red color. The bird's head is tilted slightly to the side, giving the impression of it looking regal and majestic. The background is blurred, drawing attention to the bird's striking appearance.

Jenga-3Stage / 157s

In a charming Parisian caf\u00e9, a panda sits at a quaint wooden table, sipping coffee from a delicate porcelain cup. The panda, wearing a stylish beret and a striped scarf, gazes out the window at the bustling Paris streets, where the Eiffel Tower is visible in the distance. The caf\u00e9's interior is adorned with vintage posters and warm lighting, creating a cozy ambiance. The panda's gentle movements and serene expression reflect a moment of pure contentment, as the aroma of freshly brewed coffee fills the air, blending with the soft murmur of conversations and the clinking of cups.

Jenga-3Stage / 157s

A serene suburban driveway stretches out, lined with vibrant autumn trees shedding their golden leaves. The scene begins with a close-up of the driveway's smooth, dark asphalt, glistening from a recent rain. As the camera pans out, a charming brick house with ivy climbing its walls comes into view, framed by meticulously trimmed hedges. A classic red bicycle leans against a white picket fence, adding a nostalgic touch. The driveway is bordered by colorful flower beds, with butterflies fluttering around. In the distance, a family car slowly pulls in, its headlights cutting through the early evening mist, creating a warm, inviting atmosphere.

Jenga-3Stage / 157s

A vibrant red fire hydrant stands prominently to the right of a weathered stop sign, both set against a backdrop of a quiet suburban street. The hydrant, with its glossy paint and metallic sheen, contrasts sharply with the slightly rusted, faded stop sign. The scene is framed by a row of neatly trimmed hedges and a distant view of charming houses with white picket fences. The sky above is a clear blue, with a few fluffy clouds drifting lazily. The sunlight casts gentle shadows, highlighting the textures of the hydrant and the sign, creating a picturesque and serene neighborhood moment.

Jenga-Turbo / 225s

A meticulous individual stands in a cozy, sunlit room, wearing a crisp white shirt and dark jeans, carefully ironing a freshly laundered blue dress shirt on a sleek, modern ironing board. The steam rises gently from the iron, creating a soft, hazy effect in the warm light. The room is adorned with potted plants and a large window that lets in natural light, casting a serene glow. The person\u2019s focused expression and precise movements reflect their dedication to the task. As they glide the iron smoothly over the fabric, the wrinkles disappear, leaving the shirt perfectly pressed and ready to wear.

Jenga-Turbo / 225s

A sleek, black motorcycle with chrome accents stands proudly on a winding mountain road, its polished surface gleaming under the midday sun. The camera zooms in to capture the intricate details of the engine, the leather seat, and the handlebars, showcasing the craftsmanship. The scene shifts to the motorcycle speeding along the road, the rider in a black leather jacket and helmet, leaning into a curve with the majestic mountains and a clear blue sky in the background. The roar of the engine echoes through the serene landscape, emphasizing the power and freedom of the ride. Finally, the motorcycle comes to a stop at a scenic overlook, the rider dismounting to take in the breathtaking view, the machine standing as a symbol of adventure and exploration.

Jenga-Turbo / 225s

A skilled skier, clad in a vibrant red jacket, black pants, and a matching helmet, glides effortlessly down a pristine, snow-covered mountain slope. The sun shines brightly, casting a golden glow on the untouched snow, while evergreen trees line the edges of the trail. The skier carves graceful arcs in the snow, sending up sprays of powder with each turn. In the background, majestic, snow-capped peaks rise against a clear blue sky, creating a breathtaking alpine panorama. The skier's movements are fluid and precise, embodying the thrill and freedom of the sport in this winter wonderland.

Jenga-Turbo / 225s

Gwen Stacy, in her iconic Spider-Gwen suit with a white hood and pink accents, sits cross-legged on a rooftop under a twilight sky, engrossed in a thick, leather-bound book. The cityscape behind her is bathed in the soft glow of streetlights and the distant hum of traffic. Her expressive eyes, framed by her mask, move intently across the pages, occasionally glancing up as if lost in thought. The animated style captures the fluidity of her movements, from the gentle flipping of pages to the subtle shifts in her posture. The scene transitions to a close-up of her face, revealing a serene smile as she finds solace in the story, with the vibrant colors and dynamic lines of the animation bringing her character to life.

Jenga-Turbo / 225s

A joyful Corgi with a fluffy coat and perky ears bounds through a sunlit park, the golden hues of sunset casting a warm glow on the scene. In super slow motion, the Corgi's playful leaps and bounds are captured in exquisite detail, each movement highlighting its exuberance and energy. The dog's tongue lolls out in pure delight as it chases after a fluttering leaf, its paws kicking up tiny tufts of grass. The background features tall trees with leaves gently swaying in the evening breeze, and the sky is painted in shades of orange and pink, enhancing the serene yet lively atmosphere.

Jenga-Turbo / 225s

A fluffy orange tabby cat with white paws and a bushy tail sits on a polished wooden floor, eagerly eating from a ceramic bowl decorated with fish patterns. The camera captures the cat's delicate whiskers twitching and its ears perked up, fully immersed in its meal. The sunlight streaming through a nearby window casts a warm glow on the scene, highlighting the cat's soft fur and the gentle clinking sound of kibble against the bowl. The background features a cozy kitchen setting with rustic cabinets and a potted plant, adding to the homey atmosphere.

Jenga+AccVideo / 76s

A movie trailer featuring the adventures of the 30 year old space man wearing a red wool knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colors.

Jenga+AccVideo / 76s

A close up view of a glass sphere that has a zen garden within it. There is a small dwarf in the sphere who is raking the zen garden and creating patterns in the sand.

Jenga+Wan2.1-1.3B / 24s

A Mars rover moving on Mars.

Jenga+Wan2.1-1.3B / 24s

Campfire at night in a snowy forest with starry sky in the background..

Jenga+Wan2.1-1.3B / 24s

A cat wearing sunglasses and working as a lifeguard at a pool.

Comparisons

The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it's tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.

HunyuanVideo / 1625s

TeaCache-fast / 708s

SVG / 908s

Jenga-Base / 347s

Jenga-Turbo / 225s

Jenga-3stage / 157s

A drone camera circles around a beautiful historic church built on a rocky outcropping along the Amalfi Coast, the view showcases historic and magnificent architectural details and tiered pathways and patios, waves are seen crashing against the rocks below as the view overlooks the horizon of the coastal waters and hilly landscapes of the Amalfi Coast Italy, several distant people are seen walking and enjoying vistas on patios of the dramatic ocean views, the warm glow of the afternoon sun creates a magical and romantic feeling to the scene, the view is stunning captured with beautiful photography.

HunyuanVideo / 1625s

TeaCache-fast / 708s

SVG / 908s

Jenga-Base / 347s

Jenga-Turbo / 225s

Jenga-3stage / 157s

Photorealistic closeup video of two pirate ships battling each other as they sail inside a cup of coffee.

HunyuanVideo / 1625s

TeaCache-fast / 708s

SVG / 908s

Jenga-Base / 347s

Jenga-Turbo / 225s

Jenga-3stage / 157s

A movie trailer featuring the adventures of the 30 year old space man wearing a red knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colors.

HunyuanVideo / 1625s

TeaCache-fast / 708s

SVG / 908s

Jenga-Base / 347s

Jenga-Turbo / 225s

Jenga-3stage / 157s

Ablation Study

1. Effect of different text-attention amplification bias values that affect field of views

negative bias

zero bias

low bias

mid bias

high bias

w/o bias: abnormal FOV (360P first stage)

original 1 stage (720P first stage)

bias with 2 stages (540P first stage)

bias with 3 stages (360P first stage)

w/o bias: abnormal FOV (360P first stage)

original 1 stage (720P first stage)

bias with 2 stages (540P first stage)

bias with 3 stages (360P first stage)

2. Effectiveness of the Adjacency Mask

w/o adjacency mask

with adjacency mask

w/o adjacency mask

with adjacency mask

Limitation Analysis

Please hover on the video to see the text prompt

Main failure case: latent misalignment when resizing

A. hand has wrong content

A person is clapping

A. with enhanced prompt

A person in a vibrant red sweater stands in a warmly lit room, their face beaming with joy. They begin clapping enthusiastically, their hands moving rhythmically, creating a sense of celebration. The camera captures their expressive eyes and wide smile, highlighting their genuine happiness. As they continue clapping, the background reveals a cozy living space with soft lighting, adding to the intimate and cheerful atmosphere. The sound of their claps resonates, filling the room with a sense of accomplishment and shared joy.

B. boundary misalignment

A red chair

B. clear boundary with enhanced prompt

A striking red chair sits alone in the center of a minimalist room, its vibrant color contrasting sharply with the white walls and polished wooden floor. The chair, with its sleek, modern design and plush cushioning, invites viewers to imagine the comfort it offers. Sunlight streams through a nearby window, casting soft shadows and highlighting the chair's rich hue. As the camera slowly circles around, the chair's elegant curves and fine craftsmanship become more apparent. The scene transitions to a close-up, revealing the intricate stitching on the fabric and the subtle texture that adds depth to its appearance.

Alternative Solution: Use enhanced prompts / Generate contents with complex scene & textures

Based on enhanced prompts, we can eliminate the quality degradation of the generated video, with a much smaller inital resolution (360P).

dynamic scene

Balloon full of water exploding in extreme slow motion.

contents with detailed textures

A happy fuzzy panda playing guitar nearby a campfire, snow mountain in the background.

static scene with enhanced prompt

In a still frame, a weathered stop sign stands prominently at a quiet intersection, its red paint slightly faded and edges rusted, evoking a sense of time passed. The sign is set against a backdrop of a serene suburban street, lined with tall, leafy trees whose branches gently sway in the breeze. The sky above is a soft gradient of twilight hues, transitioning from deep blue to a warm orange, suggesting the end of a peaceful day. The surrounding area is calm, with neatly trimmed lawns and quaint houses, their windows glowing softly with indoor lights, adding to the tranquil atmosphere.

complex scene with enhanced prompt

A bright, spacious classroom filled with natural light streaming through large windows, casting a warm glow on the wooden desks arranged in neat rows. The walls are adorned with colorful educational posters and a large world map, creating an inviting and stimulating environment. In the front, a cheerful teacher stands by a whiteboard, writing an engaging lesson with vibrant markers. Students of diverse backgrounds sit attentively, their faces reflecting curiosity and eagerness to learn. Some are raising their hands, eager to participate, while others are engrossed in their textbooks....

Jenga 🧱

Training-Free Efficient Video Generation via Dynamic Token Carving

Supplementary Materials

Video are compressed with ffmpeg to reduce the size of the files.

Teaser Video

Hover on the video to see corresponding text prompts

More Showcases

Hover on the video to see corresponding text prompts

Comparisons

Photorealistic closeup video of two pirate ships battling each other as they sail inside a cup of coffee.

A movie trailer featuring the adventures of the 30 year old space man wearing a red knitted motorcycle helmet, blue sky, salt desert, cinematic style, shot on 35mm film, vivid colors.

Ablation Study

1. Effect of different text-attention amplification bias values that affect field of views

2. Effectiveness of the Adjacency Mask

Limitation Analysis

Please hover on the video to see the text prompt

Main failure case: latent misalignment when resizing

Alternative Solution: Use enhanced prompts / Generate contents with complex scene & textures

Based on enhanced prompts, we can eliminate the quality degradation of the generated video, with a much smaller inital resolution (360P).